Activity Management in distributed architecture using AWS SWF - amazon-web-services

Question

I have two servers(EC2 instances). In one server(server 1) i have 3 Batch and on another(server 2) i have 4 Batch. Now, one of the batch in server 2 needs to be executed only after the successful execution of a batch in server 1.
updated
Promise<Void> r12 = null
new TryCatchFinally(){
// First server job sequencing
Promise<Void> r11 = client1.b1();
r12 = client1.b2(r11);
Promise<Void> r13 = client1.b3(r12);
Promise<Void> r14 = client1.b4(r13);
}
#Override
protected void doCatch(Throwable e) throws Throwable {
System.out.println("Failed to execute commands in server 1");
}
#Override
protected void doFinally() throws Throwable {
// cleanup
}
}
new TryCatchFinally(){
// Second server job sequencing
Promise<Void> r21 = client2.b1();
// Will execute only when both parameters are ready
Promise<Void> r22 = client2.b2(r21, r12);
Promise<Void> r23 = client2.b3(r22);
Promise<Void> r24 = client2.b4(r23);
}
#Override
protected void doCatch(Throwable e) throws Throwable {
System.out.println("Failed to execute commands in server 2");
}
#Override
protected void doFinally() throws Throwable {
// cleanup
}
}
Any of the activity in any server can throw any custom exception. But the execution of any activity in a sever should not be cancelled because of exception thrown by activity in another server. Activity in a server should only be cancelled in case one of the activity in its own server throws any exception. (Dependent activity should also gets cancelled out irrespective of server if the activity on which it is dependent fails or throw any exception). For this what I did is wrapped it into two separate try catch block.
how to terminate the Workflow Execution if the activity of server 1 and server 2 both throws any exception or fails?

Answer 1

You can wrap each Spring Batch execution into a SWF activity and then use SWF decider to sequence these activities. See AWS Flow Framework documentation and recipes for more info.
Added after reading the updated description of the problem:
You can use Promises to sequence activities in any way. So in your case I would do something like:
// First server job sequencing
Promise<Void> r11 = client1.b1();
Promise<Void> r12 = client1.b2(r11);
Promise<Void> r13 = client1.b3(r12);
Promise<Void> r14 = client1.b4(r13);
// Second server job sequencing
Promise<Void> r21 = client2.b1();
// Will execute only when both parameters are ready
Promise<Void> r22 = client2.b2(r21, r12);
Promise<Void> r23 = client2.b3(r22);
Promise<Void> r24 = client2.b4(r23);
If any of the activities throws an exception it would cancel all outstanding activities and fail the workflow unless the exception is explicitly catched and handled using TryCatchFinally. Activity that wasn't started (for example because it is waiting for its parameters of type Promise become ready) is cancelled immediately. An activity that is executing should explicitly handle cancellation. See "Activity Heartbeat" section from Error Handling page of the AWS Flow Framework Guide for more info.
Added the error handling part:
You wrap the part that shouldn't affect other parts of the worklfow in the TryCatch. So in this example any client2 activity throwing an exception cancels all future client2 activities, but not activities called on client1 as exception is not thrown into its scope.
// First server job sequencing
Promise<Void> r11 = client1.b1();
final Promise<Void> r12 = client1.b2(r11);
Promise<Void> r13 = client1.b3(r12);
Promise<Void> r14 = client1.b4(r13);
new TryCatch(){
#Override
protected void doTry() throws Throwable {
// Second server job sequencing
Promise<Void> r21 = client2.b1();
// Will execute only when both parameters are ready
Promise<Void> r22 = client2.b2(r21, r12);
Promise<Void> r23 = client2.b3(r22);
Promise<Void> r24 = client2.b4(r23);
}
#Override
protected void doCatch(Throwable e) throws Throwable {
// Handle exception without rethrowing it.
}
}